CPU & RAM burn-in guide

[fakhrain]

Does burn-in works better with slightly higher temperature?

[Jochenp]

I would guess so.
I see burning-in as overloading your cpu, forcing it to operate beyond what it would normally accept for a long period of time so it not-so-gently adjusts to it.
Much like a heavy workout bodytrainers do, except that it is almost impossible to compare the two without seeing your pc as a person

[MrSeanKon]

Announcement!
Well the XS user Jochenp replies for me better and faster of course!
Yeah fakhrain this is the main difference CPU Stretcher vs Toast and other burn-in programs:
High stress (as more as possible) but it keeps the temperatures as lower as possible!
As I read your post fakhrain you are an OcBible expert user now so you know much for it!
I am glad cos this program helped to gain some MHz!
But do not increase the Vcore too much see below for my poor Venice 3200+
Congrats!

[Jochenp]

I'm glad to help!
So you think lower temperatures actually help more than high temperatures?
I thought burning-in was more succesfull using high temperatures.
Mmm, something someone with some time on his hands should check (I wish I could)...
It probably is hard to find out what works best, since you can't undo your results and retest with another program, now can you
Jochenp

[MrSeanKon]

Jochenp this is my personal view.
I have read in the past at other forums discussion about this process.
On the other hand how we can test a CPU running two different burn-in programs???
To be more specific I also perfomed the method using Toast and CPU burn-in for San Diego 3700+ (temperatures at full load were higher) but nothing.
The CPU does not love burn-in GKR I told above it became more unstable.
I believe that a burn-in program should keep the temperature as small as possible maximizing the load cos high temperatures are dangerous.
Look for example the AMD temperature crash estimator and focus to values.
If you are stable at X MHz (a Celsius) then for (a+7) Celsius CPU will lose multiplier MHz if is not working in Breakdown Region.
But I do not want to say more about this cos here is an another thread where I analyze my AMD database.
Anyway if you have time and patience stretch your CPU as more as possible and when you reach its limits running CPU Stretcher then try an another program and see what is going on.
If you want to post info it is appreciated

An alternative opinion:

"Burn-In" works for a couple of reasons.
On current cpu's, there's a heatspreader. Under this chunk of metal is some sort of Thermal Interface Material. As this material gets hot, it softens, and penetrates into the cracks and crevices of the IHS, making contact between teh cpu and the IHS better. SO, running a cpu @ high temps facilitates this process.
This is even more apparant, at least in my own "experiments", with cpu's that are soldered, rather than simply having a paste-based TIM.
These cpu's need higher temps in order to soften the solder, and thereby benefit more from "burn-in".
I've also experimentsed with no-IHS cpu's...tossing them in the oven, and slowly raising temps, 1 degree @ a time(or that's what I had hoped for, but my oven does not have such "granularity"). In this instance, you can create better contact between the actual silicon core, and the pcb it's mounted to, by softening the solder attaching the two devices. Thankfully, because of surface tension, worries about shorting out the cpu are minimal, as long as temps don't get too high.
Your process, MrSeanKon, aids these processes. As you run through your test, trying to find a max, the cpu heats up....blah, blah, blah. This is why results must be obtained with a new cpu, and not an older one...the thermal interface material or solder has already been stressed, and can only flow into so many places.

about the

Does burn-in works better with slightly higher temperature?

[Jasy]

Tell me something...
Burning in with Sandra is good too, isn't it?

[MrSeanKon]

What can I say now........
First of all Sisoft Sandra is a complete package etc....
Yeah some people (especially for RAM) use Sandra for burn-in.......
I have not use it for this process, only for benchmarking.
Therefore you have to choose the best tool for you.
In the past I used Toast for CPU burn-in.
As you read below other readers use CPU burn for example.
I just want to add more and more capabilities in OcBible package instead of using many different programs....
But this needs time and knowledge.

[Jasy]

Before i only could get 2260mhz with stock voltage, after burn in with sandra i can push it into 2414mhz stock volt. CPU burnin has older instructions, sandra has it updated.

[MrSeanKon]

Before i only could get 2260mhz with stock voltage, after burn in with sandra i can push it into 2414mhz stock volt.

WOW 150 MHz gain!!!

CPU burnin has older instructions

Maybe...
IMHO any programmer must update the source code.

sandra has it updated.

Sandra is an up-to-date program of course.

[MrSeanKon]

Here are OcBible, Guidemania, CPU & RAM stretcher

The new OcBible v1.51 has a new interface and updates to RAM Stretcher.
Moreover an update for CPU Stretcher is available.
Both programs require Framework v2.0 (and Guidemania v1.2).

Continued:
Of course you can run together the burn-in programs.
Study the first photo.
System Reporter shows the % usage of CPU and RAM.
Thus it is not necessary to use other programs :p
Moreover it shows a taskbar for pagefile size.

New features for CPU and RAM Stretcher!
The cheat button!
When you start the programs is read = you cannot cheat (edit) the comboboxes (second photo).

If you press it then its color changes to green = LOL cheat the programs!
This is useful if you want to enter a different value for ZZZ or threads.

[giorgos th.]

RAM Stretcher loads only one thread.
Why not one more option for threads number?
If no what's your suggestion?

[MrSeanKon]

RAM Stretcher loads only one thread.

Yes you are right.

Why not one more option for threads number?

I think no reason for this.

If no what's your suggestion?

The basic idea is to stretch RAM not the CPU although the program stretches RAM & CPU (just figure out Task Manager).
Someone who has a multicore CPU he/she can load CPU Stretcher also just to push the load at 100% if he/she wants.
To be more specific:
Assume a quad core CPU.
The he/she run one RAM Stretcher (25% CPU load no ZZZ) and one CPU Stretcher instance (but three threads and ZZZ=0).
Am I clear?

[Jochenp]

Holy Threadresurrection moly
I was going through old posts when I read this thread. Does anyone have a theory on this "burn-in" phenomenon? I read somewhere that it does something on a microscopic level, widening some electrical paths or something and that it would eventually kill you cpu by first having it clock higher but that after too much pushing it would die??

[Nightstar]

Electromigration maybe?

[Jochenp]

Yes, exactly, that's what he was talking about. Could you explain it further?

[Nightstar]

I've only a little knowlege on the subject so I doubt I could explain how the phenomenon enhances IC performance. But I'll speculate

Electromigration is the transport of conductor material due to current. It's been studied mostly due to ICs that fail prematurely so it probably goes unnoticed in most PC compontents. Perhaps newly manufactured ICs have small flaws that increase their internal resistance. Perhaps electromigration removes the defects reducing the internal resistance of the IC allowing higher clocks/lower volts.

Dunno, it's idle speculation. But I imagine it being like unto a bearing. When new tiny defects reduce it's efficiency until worn off by friction. eventually(particularly if the lubricant is changed after this break-in period) efficiency reaches a peak then begins to fall off again until failure due to wear.

[RASTAVIPER]

I guess the whole idea is to stretch a hardware part as a cpu or a ram through the burnin in process and afterwards you can receive the same results with even lower voltage.
In a way, the hardware "learns" to operate in higher speed but with lower voltage, than it needed in the beginning.

[Jochenp]

That's not really how it works Electronics don't 'learn', we learn to understand them